Making sense of ranges

[Jonathan M Davis]

On Saturday, March 24, 2012 18:14:46 Ali Çehreli wrote:
> I looked for rationale on Andrei's article. There is this bit about STL
> forward iterators:
> 
> <quote>
> Input and forward iterators are syntactically identical but subtly
> different semantically—copying a forward iterator saves iteration state,
> but copying an input iterator just creates a new view of the same
> iteration state.
> </quote>
> 
> I guess that's why 'save' is explicit on ForwardRange.

Yes. If a range is a reference type rather than a value type or a dynamic 
array, then you _must_ have save in order to get a copy of its current 
state. However, since most ranges are either dynamic arrays or value types, 
functions are often written without using save like they should and would 
end up consuming a forward range if it's a reference type (e.g. a class). 
But save was created primarily to make it possible to have classes which are 
ranges rather than relying on assignment making a copy (and not all structs 
are value types, so save can be required for them as well).

- Jonathan M Davis